CRISPR-Cas9 is a versatile genome-editing technology that is widely used for studying the functionality of genetic elements, creating genetically modiﬁed organisms as well as preclinical research of genetic disorders. However, the occurrence of off-target mutations can limit its applicability. The high frequency of off-target activity-RGEN (RNA-guided endonuclease)-induced mutations at sites other than the intended on-target site, which may cause genomic instability and disrupt the functionality of otherwise normal genes, is one major concern. Because CRISPR-Cas9 causes permanent genome alterations, its off-target effects must be accurately proﬁled and controlled when applied in gene therapy. In spite of the remarkable advantages of CRISPR-based genome editing with respect to traditional therapies, safety and efficacy issues have to be fully addressed.
CRISPR-Cas9 is the state-of-the-art technology for editing and manipulating nucleic acids. Although the targeting speciﬁcity of Cas9 is believed to be tightly controlled by the 20-nt guide sequence of the sgRNA and the presence of a PAM adjacent to the target sequence in the genome, potential off-target cleavage activity could still occur on DNA sequence with even three to ﬁve base pair mismatches in the PAM-distal part of the sgRNA-guiding sequence. At the molecular level, off-target effects are the unselective cleavages of DNA sequences that do not fully match the guide RNA, bearing base pair mismatches at PAM distal sites of the DNA:RNA hybrid. Off-target sequences affect the conformational activation of the HNH nuclease domain. Off-target effects represent a severe issue hindering a full exploitation of the CRISPR-Cas9 technology in in vivo applications.
Fig 1. Crystal structure of the S. pyogenes CRISPR-Cas9 system. The target DNA (TS) and non-target DNA (NTS) strands.
Related studies reveal that base pair mismatches in the target DNA at specific distal sites with respect to the Protospacer Adjacent Motif (PAM) induce an extended opening of the RNA:DNA heteroduplex, which leads to interactions between the unwound nucleic acids and the protein counterpart. The conserved interactions between the target DNA strand and the L2 loop of the catalytic HNH domain constitute a "lock" effectively decreasing the conformational freedom of the HNH domain and its activation for cleavage. Remarkably, depending on their position at PAM distal sites, DNA mismatches leading to off-target cleavages are unable to "lock" the HNH domain, thereby identifying the ability to "lock" HNH as a key determinant.
The occurrence of off-target cleavages is directly related to the conformational state adopted by the catalytic HNH domain. Upon DNA binding, HNH undergoes a conformational change from an inactive state, in which the catalytic H840 is far away from the cleavage site on the target DNA, to an activated state that is prone for catalysis. The inactive state of the HNH domain is identified as a "conformational checkpoint" between DNA binding and cleavage, in which the RNA:DNA complementarity is recognized before the HNH domain assumes an activated conformation. Studies have shown that the presence of DNA mismatches in the DNA:RNA hybrid can trap HNH domain in the "conformational checkpoint" state, whose population increases by augmenting the number of mismatches at PAM distal sites. This poses the foundation for designing novel and more specific Cas9 variants.
Detecting off-target sites in a highly sensitive and comprehensive manner remains a key challenge in the ﬁeld of gene editing. The T7 endonuclease I assay assay suffers poor sensitivity, and it is neither practical nor cost-effective for large-scale screening. Various advanced methods for off-targeting detecting including deep sequencing (measure off-target mutations at frequencies ranging from 0.01 to 0.1%), web-based prediction tools, and ChIP-seq have been developed and widely adapted. The Web-based algorithms have an innate limitation as the tools assume that off-target sequences are closely related to the on-target site, which may miss detrimental off-target sites with less sequence similarity. ChIP-seq has also been used to identify off-target binding sites for sgRNAs complexed with catalytically inactive Cas9 (dCas9).
|T7E1 assay||Simple||Poor sensitivity, not cost-effective|
|Deep sequencing||Precise||Biased, misses potential off-target sites elsewhere in the genome|
|In silico prediction||Predicts some off-target mutation sites||Fails to predict bona-ﬁde off-target sites|
|ChIP-seq||Unbiased detection of Cas9 binding sites genome-wide||Most off-target DNA-binding sites recognized by d Cas9 are not cleaved at all by Cas9 in cells|
|GUIDE-seq||Unbiased, sensitive (0.1%), qualitative translocations, identiﬁes breakpoint hotspots||False negatives present, limited by chromatin accessibility|
|HTGTS||Identiﬁes translocations||False negatives present, limited by chromatin accessibility|
|IDLV||Programmable, sensitive (1%)||Many bona-ﬁde off-target sites cannot be captured|
|Digenome-seq||Sensitive (0.1% or lower), unbiased and cost-effective||Not widely used|
Current off-target identiﬁcation methods comprises mainly of silicon prediction and in vitro selection, which are based on the complementarity between sgRNAs and potential off-target sequences. However, silicon methods can only identify part of the off-target cleavage. And moreover, DNA binding and cleavage by Cas9 are in some cases uncoupled, that is, Cas9 can bind to but not cleave DNA sequences that are partially complementary to sgRNA and exert epigenomic regulatory effects, which is unpredictable by silicon methods relying on base pairing. So, more unbiased methods are needed to provide a comprehensive off-target proﬁle, such as genome-wide characterisation of Cas9 binding proﬁle by chromatin immunoprecipitation sequencing analysis, and genome-wide identiﬁcation of Cas9 cleavage proﬁle by GUIDE-seq.
Over the past several years, considerable efforts have been made to diminish off-target effects, and the speciﬁcity of CRISPR/Cas9 has been enhanced using some strategies. Firstly, the sgRNA sequence can be altered. sgRNA truncated by 2-3 nt are reported to reduce the off-target effects possibly because shorter sgRNA sequence has a decreased mismatch tolerance. Both the structure and composition of the guide RNA can affect the frequency of off-target effects. To select a target site that has no homologous sequence throughout the genome is a practical way for reducing off-target effects. Secondly, one potential strategy for minimizing off-target effects is to control the concentration of the Cas9-sgRNA complex by titrating the amount of Cas9 and sgRNA delivered.
Thirdly, the using of paired Cas9 nickase (a mutant form of Cas9 that generates single-stranded break rather than DSB) to generate paired nicks on the two strands of target sequence can signiﬁcantly increase target speciﬁcity because off-target single nicks are faithfully repaired. Fourthly, to further improve DNA cleavage speciﬁcity, fusions of catalytically inactive Cas9 (dCas9) with FokI nuclease domain (fCas9) have been generated. This work provides the foundation for the further characterization and improvement of Cas9 speciﬁcity and cleavage activity in vitro and in vivo. Fifth, delivery vehicles by which Cas9 and sgRNAs enter into cells can also affect on- to off-target activity ratio. Cell-penetrate-peptide-mediated delivery was reported to achieve higher on- to off-target activity ratio compared with plasmid-mediated delivery.
1. Biagioni et al. Type II CRISPR/Cas9 approach in the oncological therapy. Journal of Experimental & Clinical Cancer Research (2017) 36:80.
2. Zhang et al. Off-target Effects in CRISPR/Cas9-mediated Genome Engineering. Molecular Therapy-Nucleic Acids. 17 November 2015; 4, e264.
3. CG Ricci et al. Molecular mechanism of off-target effects in CRISPR-Cas9. bioRxiv. 19 September 2018;1-29.
4. L Xiao-Jie et al. CRISPR-Cas9: a new and promising player in gene therapy. J Med Genet. 24 February 2015; 52:289–296..
5. Cho et al. Analysis of off-target effects of CRISPR/Cas-derivedRNA-guided endonucleases and nickases. Genome Research. 2014; 24:132–141.